This jupyter notebook was implemented using PyTorch and the dataset was obtained using Kaggle cli. Platform used: Floydhub. You can get more info about U-Nets here: https://arxiv.org/abs/1505.04597.
Much of the preprocessing and basic model building was inspired by Jocic Marko's excellent starter code. Dice Loss used.
Implementation using Generative Adversarial Networks is on the way. Stay Tuned!